Andrej Karpathy has long been ahead.
He was ahead of the AI boom, having worked as a founding member of OpenAI in 2015, long before competitors like Anthropic and xAI emerged. He also got into self-driving vehicles early, steering Tesla’s autopilot effort as its head of AI.
Now, he says, “I’ve never felt this much behind as a programmer.”
In an X post on Friday, Karpathy wrote that the industry was being “dramatically refactored,” as individual programmers contribute fewer and fewer lines of code.
“I have a sense that I could be 10X more powerful if I just properly string together what has become available over the last ~year,” he wrote. “A failure to claim the boost feels decidedly like skill issue.”
I’ve never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the programmer are increasingly sparse and between. I have a sense that I could be 10X more powerful if I just properly string together what has become…
— Andrej Karpathy (@karpathy) December 26, 2025
AI has radically transformed the software engineering industry, introducing code editors like Cursor, Claude Code, and Codex, along with a slew of agentic software development tools. Business Insider’s Amanda Hoover called 2025 “the year coding changed forever.”
Karpathy was a key player in the change. In February, he coined the term “vibe coding.” To vibe code, one prompts AI to generate lines of code. (It gets its name because developers “fully give in to the vibes,” Karpathy wrote in his original post.) The Collins Dictionary named it the word of the year.
Still, Karpathy wrote that it’s like a “powerful alien tool” was handed out without a manual.
“Everyone has to figure out how to hold it and operate it, while the resulting magnitude 9 earthquake is rocking the profession,” he wrote.
In the comments, another one of the biggest names in vibe-coding agreed. Boris Cherny created Claude Code for Anthropic, now one of the most popular AI tools among developers.
Cherny wrote that he felt that way “most weeks,” and that he sometimes finds himself approaching a problem manually, not yet realizing AI can do it faster.
New graduates and early career coders may fare best in this new environment, Cherny wrote, because they don’t assume what AI can and cannot do.
“It takes significant mental work to re-adjust to what the model can do every month or two, as models continue to become better and better at coding and engineering,” he wrote.
Responding to Cherny, Karpathy wrote that he had similar experiences. He analogized the new tools to a weapon, one that sometimes “shoots pellets” or “misfires” — highlighting the work-in-progress nature of AI.
Other times, though, the tools work wonders.
“Once in a while when you hold it just right a powerful beam of laser erupts and melts your problem,” he wrote.
